4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
40 #include <lustre_acl.h>
42 #include <lustre_debug.h>
43 #include <lustre_ver.h>
44 #include <lustre_disk.h> /* for s2sbi */
45 #include <lustre_eacl.h>
47 /* for struct cl_lock_descr and struct cl_io */
48 #include <cl_object.h>
50 #include <lustre_mdc.h>
51 #include <linux/lustre_intent.h>
57 #ifndef VM_FAULT_RETRY
58 #define VM_FAULT_RETRY 0
61 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
62 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
63 #ifndef LOOKUP_CONTINUE
64 #define LOOKUP_CONTINUE LOOKUP_PARENT
67 /** Only used on client-side for indicating the tail of dir hash/offset. */
68 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
69 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
71 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
72 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
74 struct ll_dentry_data {
77 struct obd_client_handle lld_cwd_och;
78 struct obd_client_handle lld_mnt_och;
79 struct lookup_intent *lld_it;
80 unsigned int lld_sa_generation;
81 unsigned int lld_invalid:1;
82 struct rcu_head lld_rcu_head;
85 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
87 extern struct file_operations ll_pgcache_seq_fops;
89 #define LLI_INODE_MAGIC 0x111d0de5
90 #define LLI_INODE_DEAD 0xdeadd00d
92 /* remote client permission cache */
93 #define REMOTE_PERM_HASHSIZE 16
95 struct ll_getname_data {
96 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
97 struct lu_fid lgd_fid; /* target fid we are looking for */
98 int lgd_found; /* inode matched? */
101 /* llite setxid/access permission for user on remote client */
102 struct ll_remote_perm {
103 cfs_hlist_node_t lrp_list;
108 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
109 is access permission with
110 lrp_fsuid/lrp_fsgid. */
114 /* MDS has an authority for the Size-on-MDS attributes. */
115 LLIF_MDS_SIZE_LOCK = (1 << 0),
116 /* Epoch close is postponed. */
117 LLIF_EPOCH_PENDING = (1 << 1),
118 /* DONE WRITING is allowed. */
119 LLIF_DONE_WRITING = (1 << 2),
120 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
122 LLIF_SOM_DIRTY = (1 << 3),
123 /* File is contented */
124 LLIF_CONTENDED = (1 << 4),
125 /* Truncate uses server lock for this file */
126 LLIF_SRVLOCK = (1 << 5),
130 struct ll_inode_info {
131 __u32 lli_inode_magic;
135 cfs_spinlock_t lli_lock;
136 struct posix_acl *lli_posix_acl;
138 cfs_hlist_head_t *lli_remote_perms;
139 cfs_mutex_t lli_rmtperm_mutex;
141 /* identifying fields for both metadata and data stacks. */
142 struct lu_fid lli_fid;
143 /* Parent fid for accessing default stripe data on parent directory
144 * for allocating OST objects after a mknod() and later open-by-FID. */
145 struct lu_fid lli_pfid;
147 cfs_list_t lli_close_list;
148 cfs_list_t lli_oss_capas;
149 /* open count currently used by capability only, indicate whether
150 * capability needs renewal */
151 cfs_atomic_t lli_open_count;
152 struct obd_capa *lli_mds_capa;
153 cfs_time_t lli_rmtperm_time;
155 /* handle is to be sent to MDS later on done_writing and setattr.
156 * Open handle data are needed for the recovery to reconstruct
157 * the inode state on the MDS. XXX: recovery is not ready yet. */
158 struct obd_client_handle *lli_pending_och;
160 /* We need all three because every inode may be opened in different
162 struct obd_client_handle *lli_mds_read_och;
163 struct obd_client_handle *lli_mds_write_och;
164 struct obd_client_handle *lli_mds_exec_och;
165 __u64 lli_open_fd_read_count;
166 __u64 lli_open_fd_write_count;
167 __u64 lli_open_fd_exec_count;
168 /* Protects access to och pointers and their usage counters, also
169 * atomicity of check-update of lli_has_smd */
170 cfs_mutex_t lli_och_mutex;
172 struct inode lli_vfs_inode;
174 /* the most recent timestamps obtained from mds */
175 struct ost_lvb lli_lvb;
176 cfs_spinlock_t lli_agl_lock;
178 /* Try to make the d::member and f::member are aligned. Before using
179 * these members, make clear whether it is directory or not. */
183 /* serialize normal readdir and statahead-readdir. */
184 cfs_mutex_t d_readdir_mutex;
186 /* metadata statahead */
187 /* since parent-child threads can share the same @file
188 * struct, "opendir_key" is the token when dir close for
189 * case of parent exit before child -- it is me should
190 * cleanup the dir readahead. */
192 struct ll_statahead_info *d_sai;
193 struct posix_acl *d_def_acl;
194 /* protect statahead stuff. */
195 cfs_spinlock_t d_sa_lock;
196 /* "opendir_pid" is the token when lookup/revalid
197 * -- I am the owner of dir statahead. */
201 #define lli_readdir_mutex u.d.d_readdir_mutex
202 #define lli_opendir_key u.d.d_opendir_key
203 #define lli_sai u.d.d_sai
204 #define lli_def_acl u.d.d_def_acl
205 #define lli_sa_lock u.d.d_sa_lock
206 #define lli_opendir_pid u.d.d_opendir_pid
208 /* for non-directory */
210 cfs_semaphore_t f_size_sem;
211 void *f_size_sem_owner;
212 char *f_symlink_name;
215 * cfs_rw_semaphore_t {
216 * signed long count; // align u.d.d_def_acl
217 * cfs_spinlock_t wait_lock; // align u.d.d_sa_lock
218 * struct list_head wait_list;
221 cfs_rw_semaphore_t f_trunc_sem;
222 cfs_mutex_t f_write_mutex;
224 cfs_rw_semaphore_t f_glimpse_sem;
225 cfs_time_t f_glimpse_time;
226 cfs_list_t f_agl_list;
229 /* for writepage() only to communicate to fsync */
233 * whenever a process try to read/write the file, the
234 * jobid of the process will be saved here, and it'll
235 * be packed into the write PRC when flush later.
237 * so the read/write statistics for jobid will not be
238 * accurate if the file is shared by different jobs.
240 char f_jobid[JOBSTATS_JOBID_SIZE];
243 #define lli_size_sem u.f.f_size_sem
244 #define lli_size_sem_owner u.f.f_size_sem_owner
245 #define lli_symlink_name u.f.f_symlink_name
246 #define lli_maxbytes u.f.f_maxbytes
247 #define lli_trunc_sem u.f.f_trunc_sem
248 #define lli_write_mutex u.f.f_write_mutex
249 #define lli_glimpse_sem u.f.f_glimpse_sem
250 #define lli_glimpse_time u.f.f_glimpse_time
251 #define lli_agl_list u.f.f_agl_list
252 #define lli_agl_index u.f.f_agl_index
253 #define lli_async_rc u.f.f_async_rc
254 #define lli_jobid u.f.f_jobid
258 /* XXX: For following frequent used members, although they maybe special
259 * used for non-directory object, it is some time-wasting to check
260 * whether the object is directory or not before using them. On the
261 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
262 * the "ll_inode_info" size even if moving those members into u.f.
263 * So keep them out side.
265 * In the future, if more members are added only for directory,
266 * some of the following members can be moved into u.f.
269 struct cl_object *lli_clob;
271 /* mutex to request for layout lock exclusively. */
272 cfs_mutex_t lli_layout_mutex;
276 * Locking to guarantee consistency of non-atomic updates to long long i_size,
277 * consistency between file size and KMS.
279 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
282 void ll_inode_size_lock(struct inode *inode);
283 void ll_inode_size_unlock(struct inode *inode);
285 // FIXME: replace the name of this with LL_I to conform to kernel stuff
286 // static inline struct ll_inode_info *LL_I(struct inode *inode)
287 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
289 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
292 /* default to about 40meg of readahead on a given system. That much tied
293 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
294 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
296 /* default to read-ahead full files smaller than 2MB on the second read */
297 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
302 RA_STAT_DISTANT_READPAGE,
303 RA_STAT_MISS_IN_WINDOW,
304 RA_STAT_FAILED_GRAB_PAGE,
305 RA_STAT_FAILED_MATCH,
310 RA_STAT_MAX_IN_FLIGHT,
311 RA_STAT_WRONG_GRAB_PAGE,
316 cfs_atomic_t ra_cur_pages;
317 unsigned long ra_max_pages;
318 unsigned long ra_max_pages_per_file;
319 unsigned long ra_max_read_ahead_whole_pages;
322 /* ra_io_arg will be filled in the beginning of ll_readahead with
323 * ras_lock, then the following ll_read_ahead_pages will read RA
324 * pages according to this arg, all the items in this structure are
325 * counted by page index.
328 unsigned long ria_start; /* start offset of read-ahead*/
329 unsigned long ria_end; /* end offset of read-ahead*/
330 /* If stride read pattern is detected, ria_stoff means where
331 * stride read is started. Note: for normal read-ahead, the
332 * value here is meaningless, and also it will not be accessed*/
334 /* ria_length and ria_pages are the length and pages length in the
335 * stride I/O mode. And they will also be used to check whether
336 * it is stride I/O read-ahead in the read-ahead pages*/
337 unsigned long ria_length;
338 unsigned long ria_pages;
341 /* LL_HIST_MAX=32 causes an overflow */
342 #define LL_HIST_MAX 28
343 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
344 #define LL_PROCESS_HIST_MAX 10
345 struct per_process_info {
347 struct obd_histogram pp_r_hist;
348 struct obd_histogram pp_w_hist;
351 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
352 struct ll_rw_extents_info {
353 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
356 #define LL_OFFSET_HIST_MAX 100
357 struct ll_rw_process_info {
360 loff_t rw_range_start;
362 loff_t rw_last_file_pos;
364 size_t rw_smallest_extent;
365 size_t rw_largest_extent;
366 struct ll_file_data *rw_last_file;
369 enum stats_track_type {
370 STATS_TRACK_ALL = 0, /* track all processes */
371 STATS_TRACK_PID, /* track process with this pid */
372 STATS_TRACK_PPID, /* track processes with this ppid */
373 STATS_TRACK_GID, /* track processes with this gid */
377 /* flags for sbi->ll_flags */
378 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
379 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
380 #define LL_SBI_FLOCK 0x04
381 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
382 #define LL_SBI_ACL 0x10 /* support ACL */
383 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
384 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
385 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
386 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
387 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
388 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
389 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
390 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
391 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
392 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
393 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
394 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
395 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
397 /* default value for ll_sb_info->contention_time */
398 #define SBI_DEFAULT_CONTENTION_SECONDS 60
399 /* default value for lockless_truncate_enable */
400 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
401 #define RCE_HASHES 32
403 struct rmtacl_ctl_entry {
405 pid_t rce_key; /* hash key */
406 int rce_ops; /* acl operation type */
409 struct rmtacl_ctl_table {
410 cfs_spinlock_t rct_lock;
411 cfs_list_t rct_entries[RCE_HASHES];
418 pid_t ee_key; /* hash key */
419 struct lu_fid ee_fid;
420 int ee_type; /* ACL type for ACCESS or DEFAULT */
421 ext_acl_xattr_header *ee_acl;
425 cfs_spinlock_t et_lock;
426 cfs_list_t et_entries[EE_HASHES];
431 /* this protects pglist and ra_info. It isn't safe to
432 * grab from interrupt contexts */
433 cfs_spinlock_t ll_lock;
434 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
435 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
436 struct obd_uuid ll_sb_uuid;
437 struct obd_export *ll_md_exp;
438 struct obd_export *ll_dt_exp;
439 struct proc_dir_entry* ll_proc_root;
440 struct lu_fid ll_root_fid; /* root object fid */
443 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
444 struct lustre_client_ocd ll_lco;
446 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
447 struct ll_close_queue *ll_lcq;
449 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
451 struct cl_client_lru ll_lru;
453 struct lprocfs_stats *ll_ra_stats;
455 struct ll_ra_info ll_ra_info;
456 unsigned int ll_namelen;
457 struct file_operations *ll_fop;
459 /* =0 - hold lock over whole read/write
460 * >0 - max. chunk to be read/written w/o lock re-acquiring */
461 unsigned long ll_max_rw_chunk;
462 unsigned int ll_md_brw_size; /* used by readdir */
464 struct lu_site *ll_site;
465 struct cl_device *ll_cl;
467 struct ll_rw_extents_info ll_rw_extents_info;
468 int ll_extent_process_count;
469 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
470 unsigned int ll_offset_process_count;
471 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
472 unsigned int ll_rw_offset_entry_count;
473 int ll_stats_track_id;
474 enum stats_track_type ll_stats_track_type;
477 /* metadata stat-ahead */
478 unsigned int ll_sa_max; /* max statahead RPCs */
479 atomic_t ll_sa_total; /* statahead thread started
481 atomic_t ll_sa_wrong; /* statahead thread stopped for
483 atomic_t ll_agl_total; /* AGL thread started count */
485 dev_t ll_sdev_orig; /* save s_dev before assign for
487 struct rmtacl_ctl_table ll_rct;
488 struct eacl_table ll_et;
491 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
496 struct task_struct *lrr_reader;
497 cfs_list_t lrr_linkage;
501 * per file-descriptor read-ahead data.
503 struct ll_readahead_state {
504 cfs_spinlock_t ras_lock;
506 * index of the last page that read(2) needed and that wasn't in the
507 * cache. Used by ras_update() to detect seeks.
509 * XXX nikita: if access seeks into cached region, Lustre doesn't see
512 unsigned long ras_last_readpage;
514 * number of pages read after last read-ahead window reset. As window
515 * is reset on each seek, this is effectively a number of consecutive
516 * accesses. Maybe ->ras_accessed_in_window is better name.
518 * XXX nikita: window is also reset (by ras_update()) when Lustre
519 * believes that memory pressure evicts read-ahead pages. In that
520 * case, it probably doesn't make sense to expand window to
521 * PTLRPC_MAX_BRW_PAGES on the third access.
523 unsigned long ras_consecutive_pages;
525 * number of read requests after the last read-ahead window reset
526 * As window is reset on each seek, this is effectively the number
527 * on consecutive read request and is used to trigger read-ahead.
529 unsigned long ras_consecutive_requests;
531 * Parameters of current read-ahead window. Handled by
532 * ras_update(). On the initial access to the file or after a seek,
533 * window is reset to 0. After 3 consecutive accesses, window is
534 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
535 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
537 unsigned long ras_window_start, ras_window_len;
539 * Where next read-ahead should start at. This lies within read-ahead
540 * window. Read-ahead window is read in pieces rather than at once
541 * because: 1. lustre limits total number of pages under read-ahead by
542 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
543 * not covered by DLM lock.
545 unsigned long ras_next_readahead;
547 * Total number of ll_file_read requests issued, reads originating
548 * due to mmap are not counted in this total. This value is used to
549 * trigger full file read-ahead after multiple reads to a small file.
551 unsigned long ras_requests;
553 * Page index with respect to the current request, these value
554 * will not be accurate when dealing with reads issued via mmap.
556 unsigned long ras_request_index;
558 * list of struct ll_ra_read's one per read(2) call current in
559 * progress against this file descriptor. Used by read-ahead code,
560 * protected by ->ras_lock.
562 cfs_list_t ras_read_beads;
564 * The following 3 items are used for detecting the stride I/O
566 * In stride I/O mode,
567 * ...............|-----data-----|****gap*****|--------|******|....
568 * offset |-stride_pages-|-stride_gap-|
569 * ras_stride_offset = offset;
570 * ras_stride_length = stride_pages + stride_gap;
571 * ras_stride_pages = stride_pages;
572 * Note: all these three items are counted by pages.
574 unsigned long ras_stride_length;
575 unsigned long ras_stride_pages;
576 pgoff_t ras_stride_offset;
578 * number of consecutive stride request count, and it is similar as
579 * ras_consecutive_requests, but used for stride I/O mode.
580 * Note: only more than 2 consecutive stride request are detected,
581 * stride read-ahead will be enable
583 unsigned long ras_consecutive_stride_requests;
586 extern cfs_mem_cache_t *ll_file_data_slab;
587 struct lustre_handle;
588 struct ll_file_data {
589 struct ll_readahead_state fd_ras;
591 struct ccc_grouplock fd_grouplock;
594 struct file *fd_file;
595 /* Indicate whether need to report failure when close.
596 * true: failure is known, not report again.
597 * false: unknown failure, should report. */
598 bool fd_write_failed;
601 struct lov_stripe_md;
603 extern cfs_spinlock_t inode_lock;
605 extern struct proc_dir_entry *proc_lustre_fs_root;
607 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
609 return &lli->lli_vfs_inode;
613 struct inode *icbd_parent;
614 struct dentry **icbd_childp;
618 __u32 ll_i2suppgid(struct inode *i);
619 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
621 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
623 #if BITS_PER_LONG == 32
626 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
630 #define LLAP_MAGIC 98764321
632 extern cfs_mem_cache_t *ll_async_page_slab;
633 extern size_t ll_async_page_slab_size;
635 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
636 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
637 struct ll_ra_read *ll_ra_read_get(struct file *f);
639 /* llite/lproc_llite.c */
641 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
642 struct super_block *sb, char *osc, char *mdc);
643 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
644 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
645 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
647 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
648 struct super_block *sb, char *osc, char *mdc){return 0;}
649 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
650 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
651 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
653 memset(lvars, 0, sizeof(*lvars));
659 void ll_release_page(struct page *page, int remove);
660 extern struct file_operations ll_dir_operations;
661 extern struct inode_operations ll_dir_inode_operations;
662 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
663 struct ll_dir_chain *chain);
664 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
667 int ll_get_mdt_idx(struct inode *inode);
669 int ll_objects_destroy(struct ptlrpc_request *request,
671 struct inode *ll_iget(struct super_block *sb, ino_t hash,
672 struct lustre_md *lic);
673 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
674 void *data, int flag);
675 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
677 int ll_lookup_it_finish(struct ptlrpc_request *request,
678 struct lookup_intent *it, void *data);
679 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
682 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
683 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
684 int ll_writepage(struct page *page, struct writeback_control *wbc);
685 int ll_writepages(struct address_space *, struct writeback_control *wbc);
686 void ll_removepage(struct page *page);
687 int ll_readpage(struct file *file, struct page *page);
688 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
689 int ll_file_punch(struct inode *, loff_t, int);
690 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
691 void ll_clear_file_contended(struct inode*);
692 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
693 int ll_readahead(const struct lu_env *env, struct cl_io *io,
694 struct ll_readahead_state *ras, struct address_space *mapping,
695 struct cl_page_list *queue, int flags);
698 extern struct file_operations ll_file_operations;
699 extern struct file_operations ll_file_operations_flock;
700 extern struct file_operations ll_file_operations_noflock;
701 extern struct inode_operations ll_file_inode_operations;
702 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
704 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
705 ldlm_mode_t l_req_mode);
706 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
707 struct lustre_handle *lockh);
708 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
710 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
711 int ll_file_open(struct inode *inode, struct file *file);
712 int ll_file_release(struct inode *inode, struct file *file);
713 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
714 struct lov_stripe_md *lsm, lstat_t *st);
715 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
716 int ll_local_open(struct file *file,
717 struct lookup_intent *it, struct ll_file_data *fd,
718 struct obd_client_handle *och);
719 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
720 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
722 int ll_md_real_close(struct inode *inode, int flags);
723 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
724 struct obd_client_handle **och, unsigned long flags);
725 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
726 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
727 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
728 __u64 ioepoch, int sync);
729 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
730 struct md_open_data **mod);
731 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
732 struct lustre_handle *fh);
733 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
734 struct ll_file_data *file, loff_t pos,
735 size_t count, int rw);
736 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
737 struct lookup_intent *it, struct kstat *stat);
738 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
739 struct ll_file_data *ll_file_data_get(void);
740 struct posix_acl * ll_get_acl(struct inode *inode, int type);
742 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
743 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
745 # ifndef HAVE_INODE_PERMISION_2ARGS
746 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
748 int ll_inode_permission(struct inode *inode, int mask);
752 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
753 int flags, struct lov_user_md *lum,
755 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
756 struct lov_mds_md **lmm, int *lmm_size,
757 struct ptlrpc_request **request);
758 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
760 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
761 int *lmm_size, struct ptlrpc_request **request);
762 #ifdef HAVE_FILE_FSYNC_4ARGS
763 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
764 #elif defined(HAVE_FILE_FSYNC_2ARGS)
765 int ll_fsync(struct file *file, int data);
767 int ll_fsync(struct file *file, struct dentry *dentry, int data);
769 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
771 int ll_merge_lvb(struct inode *inode);
772 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
773 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
774 int ll_fid2path(struct inode *inode, void *arg);
778 int ll_dops_init(struct dentry *de, int block, int init_sa);
779 extern struct dentry_operations ll_d_ops;
780 void ll_intent_drop_lock(struct lookup_intent *);
781 void ll_intent_release(struct lookup_intent *);
782 void ll_invalidate_aliases(struct inode *);
783 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
784 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
785 #ifdef HAVE_D_COMPARE_7ARGS
786 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
787 const struct dentry *dentry, const struct inode *inode,
788 unsigned int len, const char *str, const struct qstr *d_name);
790 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
792 int ll_revalidate_it_finish(struct ptlrpc_request *request,
793 struct lookup_intent *it, struct dentry *de);
795 /* llite/llite_lib.c */
796 extern struct super_operations lustre_super_operations;
798 char *ll_read_opt(const char *opt, char *data);
799 void ll_lli_init(struct ll_inode_info *lli);
800 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
801 void ll_put_super(struct super_block *sb);
802 void ll_kill_super(struct super_block *sb);
803 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
804 void ll_clear_inode(struct inode *inode);
805 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
806 int ll_setattr(struct dentry *de, struct iattr *attr);
807 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
808 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
809 __u64 max_age, __u32 flags);
810 void ll_update_inode(struct inode *inode, struct lustre_md *md);
811 void ll_read_inode2(struct inode *inode, void *opaque);
812 void ll_delete_inode(struct inode *inode);
813 int ll_iocontrol(struct inode *inode, struct file *file,
814 unsigned int cmd, unsigned long arg);
815 int ll_flush_ctx(struct inode *inode);
816 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
817 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
819 void ll_umount_begin(struct super_block *sb);
821 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
822 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
823 void ll_dirty_page_discard_warn(cfs_page_t *page, int ioret);
824 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
825 struct super_block *);
826 void lustre_dump_dentry(struct dentry *, int recur);
827 void lustre_dump_inode(struct inode *);
828 int ll_obd_statfs(struct inode *inode, void *arg);
829 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
830 int ll_process_config(struct lustre_cfg *lcfg);
831 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
832 struct inode *i1, struct inode *i2,
833 const char *name, int namelen,
834 int mode, __u32 opc, void *data);
835 void ll_finish_md_op_data(struct md_op_data *op_data);
836 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
837 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
839 /* llite/llite_nfs.c */
840 extern struct export_operations lustre_export_operations;
841 __u32 get_uuid2int(const char *name, int len);
843 /* llite/special.c */
844 extern struct inode_operations ll_special_inode_operations;
845 extern struct file_operations ll_special_chr_inode_fops;
846 extern struct file_operations ll_special_chr_file_fops;
847 extern struct file_operations ll_special_blk_inode_fops;
848 extern struct file_operations ll_special_fifo_inode_fops;
849 extern struct file_operations ll_special_fifo_file_fops;
850 extern struct file_operations ll_special_sock_inode_fops;
852 /* llite/symlink.c */
853 extern struct inode_operations ll_fast_symlink_inode_operations;
855 /* llite/llite_close.c */
856 struct ll_close_queue {
857 cfs_spinlock_t lcq_lock;
859 cfs_waitq_t lcq_waitq;
860 cfs_completion_t lcq_comp;
861 cfs_atomic_t lcq_stop;
864 struct ccc_object *cl_inode2ccc(struct inode *inode);
867 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
868 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
870 /* specific achitecture can implement only part of this list */
871 enum vvp_io_subtype {
874 /** io called from .sendfile */
876 /** io started from splice_{read|write} */
883 enum vvp_io_subtype cui_io_subtype;
887 read_actor_t cui_actor;
891 struct pipe_inode_info *cui_pipe;
892 unsigned int cui_flags;
894 struct vvp_fault_io {
896 * Inode modification time that is checked across DLM
900 struct vm_area_struct *ft_vma;
902 * locked page returned from vvp_io
904 cfs_page_t *ft_vmpage;
905 #ifndef HAVE_VM_OP_FAULT
906 struct vm_nopage_api {
908 * Virtual address at which fault occurred.
910 unsigned long ft_address;
912 * Fault type, as to be supplied to
918 struct vm_fault_api {
922 struct vm_fault *ft_vmf;
924 * fault API used bitflags for return code.
926 unsigned int ft_flags;
932 * Read-ahead state used by read and page-fault IO contexts.
934 struct ll_ra_read cui_bead;
936 * Set when cui_bead has been initialized.
938 int cui_ra_window_set;
940 * Partially truncated page, that vvp_io_trunc_start() keeps locked
943 struct cl_page *cui_partpage;
947 * IO arguments for various VFS I/O interfaces.
950 /** normal/sendfile/splice */
951 enum vvp_io_subtype via_io_subtype;
955 #ifndef HAVE_FILE_WRITEV
956 struct kiocb *via_iocb;
958 struct iovec *via_iov;
959 unsigned long via_nrsegs;
962 read_actor_t via_actor;
966 struct pipe_inode_info *via_pipe;
967 unsigned int via_flags;
972 struct ll_cl_context {
974 struct cl_io *lcc_io;
975 struct cl_page *lcc_page;
976 struct lu_env *lcc_env;
981 struct vvp_thread_info {
982 struct ost_lvb vti_lvb;
983 struct cl_2queue vti_queue;
984 struct iovec vti_local_iov;
985 struct vvp_io_args vti_args;
986 struct ra_io_arg vti_ria;
987 struct kiocb vti_kiocb;
988 struct ll_cl_context vti_io_ctx;
991 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
993 extern struct lu_context_key vvp_key;
994 struct vvp_thread_info *info;
996 info = lu_context_key_get(&env->le_ctx, &vvp_key);
997 LASSERT(info != NULL);
1001 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1002 enum vvp_io_subtype type)
1004 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1006 ret->via_io_subtype = type;
1011 struct vvp_session {
1012 struct vvp_io vs_ios;
1015 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1017 extern struct lu_context_key vvp_session_key;
1018 struct vvp_session *ses;
1020 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1021 LASSERT(ses != NULL);
1025 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1027 return &vvp_env_session(env)->vs_ios;
1030 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1031 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1032 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1034 /* llite/llite_mmap.c */
1035 typedef struct rb_root rb_root_t;
1036 typedef struct rb_node rb_node_t;
1038 struct ll_lock_tree_node;
1039 struct ll_lock_tree {
1041 cfs_list_t lt_locked_list;
1042 struct ll_file_data *lt_fd;
1045 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1046 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1047 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1048 __u64 end, ldlm_mode_t mode);
1049 void policy_from_vma(ldlm_policy_data_t *policy,
1050 struct vm_area_struct *vma, unsigned long addr, size_t count);
1051 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1054 static inline void ll_invalidate_page(struct page *vmpage)
1056 struct address_space *mapping = vmpage->mapping;
1057 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1059 LASSERT(PageLocked(vmpage));
1060 if (mapping == NULL)
1063 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1064 truncate_complete_page(mapping, vmpage);
1067 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1069 /* don't need an addref as the sb_info should be holding one */
1070 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1072 return ll_s2sbi(sb)->ll_dt_exp;
1075 /* don't need an addref as the sb_info should be holding one */
1076 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1078 return ll_s2sbi(sb)->ll_md_exp;
1081 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1083 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1089 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1090 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1092 return ll_s2sbi(inode->i_sb);
1095 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1097 return ll_s2dtexp(inode->i_sb);
1100 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1102 return ll_s2mdexp(inode->i_sb);
1105 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1108 LASSERT(inode != NULL);
1109 fid = &ll_i2info(inode)->lli_fid;
1110 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1114 static inline int ll_mds_max_easize(struct super_block *sb)
1116 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1119 static inline __u64 ll_file_maxbytes(struct inode *inode)
1121 return ll_i2info(inode)->lli_maxbytes;
1125 int ll_setxattr(struct dentry *dentry, const char *name,
1126 const void *value, size_t size, int flags);
1127 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1128 void *buffer, size_t size);
1129 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1130 int ll_removexattr(struct dentry *dentry, const char *name);
1132 /* llite/remote_perm.c */
1133 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1134 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1136 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1137 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1138 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1139 int lustre_check_remote_perm(struct inode *inode, int mask);
1141 /* llite/llite_capa.c */
1142 extern cfs_timer_t ll_capa_timer;
1144 int ll_capa_thread_start(void);
1145 void ll_capa_thread_stop(void);
1146 void ll_capa_timer_callback(unsigned long unused);
1148 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1149 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1151 void ll_capa_open(struct inode *inode);
1152 void ll_capa_close(struct inode *inode);
1154 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1155 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1157 void ll_truncate_free_capa(struct obd_capa *ocapa);
1158 void ll_clear_inode_capas(struct inode *inode);
1159 void ll_print_capa_stat(struct ll_sb_info *sbi);
1161 /* llite/llite_cl.c */
1162 extern struct lu_device_type vvp_device_type;
1165 * Common IO arguments for various VFS I/O interfaces.
1167 int cl_sb_init(struct super_block *sb);
1168 int cl_sb_fini(struct super_block *sb);
1169 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1170 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1172 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1173 struct ll_readahead_state *ras, unsigned long index,
1175 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1176 int ll_is_file_contended(struct file *file);
1177 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1179 /* llite/llite_rmtacl.c */
1180 #ifdef CONFIG_FS_POSIX_ACL
1181 obd_valid rce_ops2valid(int ops);
1182 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1183 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1184 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1185 void rct_init(struct rmtacl_ctl_table *rct);
1186 void rct_fini(struct rmtacl_ctl_table *rct);
1188 void ee_free(struct eacl_entry *ee);
1189 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1190 ext_acl_xattr_header *header);
1191 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1192 struct lu_fid *fid, int type);
1193 void et_search_free(struct eacl_table *et, pid_t key);
1194 void et_init(struct eacl_table *et);
1195 void et_fini(struct eacl_table *et);
1200 #define LL_SA_RPC_MIN 2
1201 #define LL_SA_RPC_DEF 32
1202 #define LL_SA_RPC_MAX 8192
1204 #define LL_SA_CACHE_BIT 5
1205 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1206 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1208 /* per inode struct, for dir only */
1209 struct ll_statahead_info {
1210 struct inode *sai_inode;
1211 cfs_atomic_t sai_refcount; /* when access this struct, hold
1213 unsigned int sai_generation; /* generation for statahead */
1214 unsigned int sai_max; /* max ahead of lookup */
1215 __u64 sai_sent; /* stat requests sent count */
1216 __u64 sai_replied; /* stat requests which received
1218 __u64 sai_index; /* index of statahead entry */
1219 __u64 sai_index_wait; /* index of entry which is the
1220 * caller is waiting for */
1221 __u64 sai_hit; /* hit count */
1222 __u64 sai_miss; /* miss count:
1223 * for "ls -al" case, it includes
1224 * hidden dentry miss;
1225 * for "ls -l" case, it does not
1226 * include hidden dentry miss.
1227 * "sai_miss_hidden" is used for
1230 unsigned int sai_consecutive_miss; /* consecutive miss */
1231 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1232 * is not a hidden one */
1233 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1234 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1236 sai_in_readpage:1,/* statahead is in readdir()*/
1237 sai_agl_valid:1;/* AGL is valid for the dir */
1238 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1239 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1240 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1241 cfs_list_t sai_entries_sent; /* entries sent out */
1242 cfs_list_t sai_entries_received; /* entries returned */
1243 cfs_list_t sai_entries_stated; /* entries stated */
1244 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1245 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1246 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1247 cfs_atomic_t sai_cache_count; /* entry count in cache */
1250 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1252 void ll_stop_statahead(struct inode *dir, void *key);
1254 static inline int ll_glimpse_size(struct inode *inode)
1256 struct ll_inode_info *lli = ll_i2info(inode);
1259 cfs_down_read(&lli->lli_glimpse_sem);
1260 rc = cl_glimpse_size(inode);
1261 lli->lli_glimpse_time = cfs_time_current();
1262 cfs_up_read(&lli->lli_glimpse_sem);
1267 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1269 struct ll_inode_info *lli = ll_i2info(dir);
1270 struct ll_statahead_info *sai = lli->lli_sai;
1271 struct ll_dentry_data *ldd = ll_d2d(dentry);
1273 /* not the same process, don't mark */
1274 if (lli->lli_opendir_pid != cfs_curproc_pid())
1277 if (sai != NULL && ldd != NULL)
1278 ldd->lld_sa_generation = sai->sai_generation;
1282 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1284 struct ll_inode_info *lli;
1285 struct ll_dentry_data *ldd;
1287 if (ll_i2sbi(dir)->ll_sa_max == 0)
1290 lli = ll_i2info(dir);
1291 /* not the same process, don't statahead */
1292 if (lli->lli_opendir_pid != cfs_curproc_pid())
1295 ldd = ll_d2d(*dentryp);
1297 * When stats a dentry, the system trigger more than once "revalidate"
1298 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1299 * Under patchless client mode, the operation intent is not accurate,
1300 * which maybe misguide the statahead thread. For example:
1301 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1302 * have the same operation intent -- "IT_GETATTR".
1303 * In fact, one dentry should has only one chance to interact with the
1304 * statahead thread, otherwise the statahead windows will be confused.
1305 * The solution is as following:
1306 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1307 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1308 * will bypass interacting with statahead thread for checking:
1309 * "lld_sa_generation == lli_sai->sai_generation"
1311 if (ldd && lli->lli_sai &&
1312 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1315 return do_statahead_enter(dir, dentryp, only_unplug);
1318 /* llite ioctl register support rountine */
1325 #define LLIOC_MAX_CMD 256
1328 * Rules to write a callback function:
1331 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1332 * returned to ll_iocontrol_register. Callback functions should use this
1333 * data to check the potential collasion of ioctl cmd. If collasion is
1334 * found, callback function should return LLIOC_CONT.
1335 * @rcp: The result of ioctl command.
1338 * If @magic matches the pointer returned by ll_iocontrol_data, the
1339 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1341 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1342 struct file *file, unsigned int cmd, unsigned long arg,
1343 void *magic, int *rcp);
1345 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1346 unsigned int cmd, unsigned long arg, int *rcp);
1348 /* export functions */
1349 /* Register ioctl block dynamatically for a regular file.
1351 * @cmd: the array of ioctl command set
1352 * @count: number of commands in the @cmd
1353 * @cb: callback function, it will be called if an ioctl command is found to
1354 * belong to the command list @cmd.
1357 * A magic pointer will be returned if success;
1358 * otherwise, NULL will be returned.
1360 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1361 void ll_iocontrol_unregister(void *magic);
1365 /* lclient compat stuff */
1366 #define cl_inode_info ll_inode_info
1367 #define cl_i2info(info) ll_i2info(info)
1368 #define cl_inode_mode(inode) ((inode)->i_mode)
1369 #define cl_i2sbi ll_i2sbi
1371 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1372 const struct iattr *attr)
1374 LASSERT(attr->ia_valid & ATTR_FILE);
1375 return LUSTRE_FPRIVATE(attr->ia_file);
1378 static inline void cl_isize_lock(struct inode *inode)
1380 ll_inode_size_lock(inode);
1383 static inline void cl_isize_unlock(struct inode *inode)
1385 ll_inode_size_unlock(inode);
1388 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1390 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1391 i_size_write(inode, kms);
1394 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1396 ll_inode_size_lock(inode);
1397 i_size_write(inode, kms);
1398 ll_inode_size_unlock(inode);
1401 #define cl_isize_read(inode) i_size_read(inode)
1403 static inline int cl_merge_lvb(struct inode *inode)
1405 return ll_merge_lvb(inode);
1408 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1409 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1410 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1412 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1414 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1415 enum cl_fsync_mode mode);
1417 /** direct write pages */
1418 struct ll_dio_pages {
1419 /** page array to be written. we don't support
1420 * partial pages except the last one. */
1421 struct page **ldp_pages;
1422 /* offset of each page */
1423 loff_t *ldp_offsets;
1424 /** if ldp_offsets is NULL, it means a sequential
1425 * pages to be written, then this is the file offset
1426 * of the * first page. */
1427 loff_t ldp_start_offset;
1428 /** how many bytes are to be written. */
1430 /** # of pages in the array. */
1434 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1437 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1440 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1443 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1444 int rw, struct inode *inode,
1445 struct ll_dio_pages *pv);
1447 static inline int ll_file_nolock(const struct file *file)
1449 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1450 struct inode *inode = file->f_dentry->d_inode;
1452 LASSERT(fd != NULL);
1453 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1454 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1457 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1458 struct lookup_intent *it, __u64 *bits)
1460 if (!it->d.lustre.it_lock_set) {
1461 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1462 inode, inode->i_ino, inode->i_generation);
1463 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1464 inode, &it->d.lustre.it_lock_bits);
1465 it->d.lustre.it_lock_set = 1;
1469 *bits = it->d.lustre.it_lock_bits;
1472 static inline void ll_lock_dcache(struct inode *inode)
1474 #ifdef HAVE_DCACHE_LOCK
1475 spin_lock(&dcache_lock);
1477 spin_lock(&inode->i_lock);
1481 static inline void ll_unlock_dcache(struct inode *inode)
1483 #ifdef HAVE_DCACHE_LOCK
1484 spin_unlock(&dcache_lock);
1486 spin_unlock(&inode->i_lock);
1490 static inline int d_lustre_invalid(const struct dentry *dentry)
1492 struct ll_dentry_data *lld = ll_d2d(dentry);
1494 return (lld == NULL) || lld->lld_invalid;
1497 static inline void __d_lustre_invalidate(struct dentry *dentry)
1499 struct ll_dentry_data *lld = ll_d2d(dentry);
1502 lld->lld_invalid = 1;
1506 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1507 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1508 * else dput() of the last refcount will unhash this dentry and kill it.
1510 static inline void d_lustre_invalidate(struct dentry *dentry)
1512 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1513 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1514 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1516 spin_lock(&dentry->d_lock);
1517 __d_lustre_invalidate(dentry);
1518 if (d_refcount(dentry) == 0)
1520 spin_unlock(&dentry->d_lock);
1523 static inline void d_lustre_revalidate(struct dentry *dentry)
1525 spin_lock(&dentry->d_lock);
1526 LASSERT(ll_d2d(dentry) != NULL);
1527 ll_d2d(dentry)->lld_invalid = 0;
1528 spin_unlock(&dentry->d_lock);
1531 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1532 /* Compatibility for old (1.8) compiled userspace quota code */
1533 struct if_quotactl_18 {
1538 struct obd_dqinfo qc_dqinfo;
1539 struct obd_dqblk qc_dqblk;
1541 struct obd_uuid obd_uuid;
1543 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1544 /* End compatibility for old (1.8) compiled userspace quota code */
1546 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1547 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1549 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1550 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1552 #endif /* LLITE_INTERNAL_H */